WEBVTT 00:00:00.000 --> 00:00:03.600 Okay, so imagine this. Imagine posting professional, 00:00:03.799 --> 00:00:07.839 really engaging videos every single day across 00:00:07.839 --> 00:00:10.539 all the big social platforms. Right. But you're 00:00:10.539 --> 00:00:13.259 never actually filming. You're not editing. You 00:00:13.259 --> 00:00:15.019 don't even have to speak into a microphone. This 00:00:15.019 --> 00:00:17.280 whole thing, this whole content operation just 00:00:17.280 --> 00:00:19.460 runs by itself. It generates everything while 00:00:19.460 --> 00:00:21.359 you sleep. It definitely sounds like science 00:00:21.359 --> 00:00:24.260 fiction, but it's absolutely possible right now. 00:00:24.320 --> 00:00:26.679 And here's the really revolutionary part. The 00:00:26.679 --> 00:00:29.160 sources we looked at... suggest you can generate 00:00:29.160 --> 00:00:31.800 something like what 150 professional clips a 00:00:31.800 --> 00:00:34.880 month a huge amount of content yeah and it can 00:00:34.880 --> 00:00:38.140 cost you around 145 dollars total that breaks 00:00:38.140 --> 00:00:41.079 down to i mean less than 50 cents for each finished 00:00:41.079 --> 00:00:43.759 video that just completely changes the economics 00:00:43.759 --> 00:00:46.299 of content it's a whole new dimension so welcome 00:00:46.299 --> 00:00:49.000 to the deep dive Today, we are unpacking the 00:00:49.000 --> 00:00:51.380 complete blueprint for this kind of automated 00:00:51.380 --> 00:00:54.439 AI avatar system. We're not talking about, you 00:00:54.439 --> 00:00:57.119 know, general ideas here. We're going to dissect 00:00:57.119 --> 00:00:59.820 the specific tools, the really crucial technical 00:00:59.820 --> 00:01:03.200 stuff, and the step -by -step workflow. And our 00:01:03.200 --> 00:01:05.920 mission here is to give you a quick but still 00:01:05.920 --> 00:01:08.939 very thorough understanding of how to build this, 00:01:09.019 --> 00:01:12.359 this low cost, high volume pipeline. We're going 00:01:12.359 --> 00:01:14.099 to start with the proof that it actually works 00:01:14.099 --> 00:01:16.540 and, you know, why audiences are okay with it. 00:01:16.920 --> 00:01:19.519 Then we have to get into the core architecture, 00:01:19.760 --> 00:01:22.780 especially that non -negotiable step of self 00:01:22.780 --> 00:01:25.060 -hosting. And finally, we'll walk through the 00:01:25.060 --> 00:01:27.840 whole process from scraping a viral idea all 00:01:27.840 --> 00:01:29.819 the way to hitting publish. So let's jump in. 00:01:29.879 --> 00:01:32.680 Let's do it. The vision of automating video creation 00:01:32.680 --> 00:01:36.900 with zero manual input is so compelling. It's 00:01:36.900 --> 00:01:39.959 the ultimate passive content dream. It is. But 00:01:39.959 --> 00:01:41.659 for something like this to be a real business 00:01:41.659 --> 00:01:44.599 model, you need proof. You need proof that audiences 00:01:44.599 --> 00:01:48.379 actually accept 100 % AI -generated video. And 00:01:48.379 --> 00:01:50.719 they do. The data is pretty clear on this. There 00:01:50.719 --> 00:01:52.560 are multiple creators who are winning with this 00:01:52.560 --> 00:01:54.659 strategy right now. The sources point to one 00:01:54.659 --> 00:01:57.239 educational creator, has about 60 ,000 followers, 00:01:57.459 --> 00:01:59.500 and they're consistently getting six -figure 00:01:59.500 --> 00:02:02.379 view counts. Wow. And every single piece of content 00:02:02.379 --> 00:02:05.700 is 100 % AI. And that brings up a really important 00:02:05.700 --> 00:02:07.959 question about audience tolerance. Because if 00:02:07.959 --> 00:02:10.060 you look really closely at some of these, the 00:02:10.060 --> 00:02:12.560 lip sync can be a little off. It can be. The 00:02:12.560 --> 00:02:16.120 movement might feel a bit robotic. And yet the 00:02:16.120 --> 00:02:18.379 videos still get thousands of likes, thousands 00:02:18.379 --> 00:02:21.870 of shares. So why? That is the critical insight 00:02:21.870 --> 00:02:25.110 right there. The audience has kind of shifted 00:02:25.110 --> 00:02:28.330 its priorities. They forgive these minor technical 00:02:28.330 --> 00:02:31.370 flaws, you know, the imperfect lip sync. If the 00:02:31.370 --> 00:02:34.409 information underneath it all delivers real concentrated 00:02:34.409 --> 00:02:36.969 value. So the content itself. Content quality, 00:02:37.330 --> 00:02:41.150 the insight, the aha moment that trumps avatar 00:02:41.150 --> 00:02:44.050 perfection every single time. We see that with 00:02:44.050 --> 00:02:46.009 creators like Sky Generated. They're making educational 00:02:46.009 --> 00:02:49.870 content about AI tools. Yeah. by using AI tools 00:02:49.870 --> 00:02:52.729 to do it. It's so meta. It is, and it's successful. 00:02:52.949 --> 00:02:56.650 They've got videos with over 857 ,000 views because 00:02:56.650 --> 00:02:58.610 the information is just what people are searching 00:02:58.610 --> 00:03:00.250 for. And then you have the hybrid structure, 00:03:00.430 --> 00:03:02.610 which I think is just a brilliant strategic move. 00:03:02.750 --> 00:03:04.789 You look at Rowan Chung's model, which is super 00:03:04.789 --> 00:03:07.830 successful. The AI avatar does the intro, right? 00:03:08.430 --> 00:03:10.789 introducing the main idea, and then boom, the 00:03:10.789 --> 00:03:13.430 video cuts straight to engaging B -roll footage 00:03:13.430 --> 00:03:15.689 or screen recordings. That seems like it solves 00:03:15.689 --> 00:03:17.629 the attention problem. It gives you the consistent 00:03:17.629 --> 00:03:20.110 presenter, the brand face, but you don't need 00:03:20.110 --> 00:03:22.030 the avatar to hold the screen for the whole clip. 00:03:22.050 --> 00:03:24.090 Exactly. Which is where you start to notice those 00:03:24.090 --> 00:03:26.270 little imperfections. Exactly right. You use 00:03:26.270 --> 00:03:28.490 the avatar for brand recognition and just sheer 00:03:28.490 --> 00:03:31.870 volume, but you rely on other high -quality visuals 00:03:31.870 --> 00:03:34.689 to keep people engaged. It's just smart. So it's 00:03:34.689 --> 00:03:37.740 a strategic choice then. Hybrid versus 100 % 00:03:37.740 --> 00:03:41.000 AI really just depends on if you're prioritizing 00:03:41.000 --> 00:03:44.659 raw output or maybe a deeper audience connection. 00:03:45.000 --> 00:03:47.500 It's about consistency and how fast you can deploy. 00:03:47.879 --> 00:03:50.620 Okay. So let's unpack the structure of this system. 00:03:50.699 --> 00:03:53.300 We're talking about three different automation 00:03:53.300 --> 00:03:55.460 loops all working together. So this is not just 00:03:55.460 --> 00:03:58.500 one tool. It's an integrated machine. That's 00:03:58.500 --> 00:04:01.319 right. So first you have the AI contact creator. 00:04:01.479 --> 00:04:03.419 That's what handles the heavy lifting script 00:04:03.419 --> 00:04:06.860 generation, voiceover, the avatar animation and 00:04:06.860 --> 00:04:08.740 putting it all together. Second, you've got the 00:04:08.740 --> 00:04:11.439 automatic publisher that takes the finished video 00:04:11.439 --> 00:04:13.819 and it schedules it, distributes it across TikTok, 00:04:14.099 --> 00:04:16.779 YouTube Shorts, Instagram Reels, all of it. And 00:04:16.779 --> 00:04:19.019 then the third loop is the fuel. That's the automatic 00:04:19.019 --> 00:04:22.480 idea scraper, which is monitoring places like 00:04:22.480 --> 00:04:25.040 X constantly looking for new viral content. So 00:04:25.040 --> 00:04:27.139 the system just never runs out of ideas. Right. 00:04:27.610 --> 00:04:30.110 And this whole machine, which, you know, it sounds 00:04:30.110 --> 00:04:32.930 really complicated, it actually runs on two core 00:04:32.930 --> 00:04:35.990 pieces of software that act like a central nervous 00:04:35.990 --> 00:04:39.189 system. The control center is Airtable. That's 00:04:39.189 --> 00:04:41.470 where you keep your ideas, manage your avatars, 00:04:41.470 --> 00:04:44.069 you track the status of every single video from 00:04:44.069 --> 00:04:46.529 idea to publish. And what's the automation engine 00:04:46.529 --> 00:04:49.480 itself? That would be N8n. If you haven't used 00:04:49.480 --> 00:04:52.480 it, N8n is automation software. It connects all 00:04:52.480 --> 00:04:54.360 these different AI tools together, kind of like 00:04:54.360 --> 00:04:56.620 data Lego blocks. It lets you build these incredibly 00:04:56.620 --> 00:04:59.220 complex workflows. Right. Connecting 11 labs, 00:04:59.480 --> 00:05:01.959 chat GPT, all your video tools without writing 00:05:01.959 --> 00:05:04.060 a single line of code. And this is where it gets 00:05:04.060 --> 00:05:07.199 really, really important because the guide points 00:05:07.199 --> 00:05:09.720 out a critical technical requirement, one that 00:05:09.720 --> 00:05:11.759 just kills the whole pipeline if you get it wrong. 00:05:11.879 --> 00:05:15.740 Yes. The system absolutely must use self -hosted 00:05:15.740 --> 00:05:19.360 N8n. This is the non -negotiable. 100%. If you 00:05:19.360 --> 00:05:22.120 try to use the standard N8N cloud, you just run 00:05:22.120 --> 00:05:25.399 into these major file handling problems. Essential 00:05:25.399 --> 00:05:27.339 video processing functions, like being able to 00:05:27.339 --> 00:05:30.680 write a file to a disk or run FFmpeg, they just 00:05:30.680 --> 00:05:32.920 won't work. Okay, you said FFmpeg. For people 00:05:32.920 --> 00:05:35.050 who aren't, you know... deep into video software. 00:05:35.290 --> 00:05:37.589 What is that exactly and why does it need that 00:05:37.589 --> 00:05:40.850 special access? So FFmpeg is basically the engine 00:05:40.850 --> 00:05:42.870 that handles all the video processing. It's the 00:05:42.870 --> 00:05:45.810 thing that cuts, resizes crops and stitches those 00:05:45.810 --> 00:05:48.050 video parts together. Got it. And because the 00:05:48.050 --> 00:05:50.009 automation is physically changing these large 00:05:50.009 --> 00:05:52.949 video files, it needs direct access to the server's 00:05:52.949 --> 00:05:55.529 hard drive. Standard cloud setups block that 00:05:55.529 --> 00:05:59.050 for security reasons. So self -hosting is, well, 00:05:59.149 --> 00:06:00.930 it's a bit of an administrative headache, isn't 00:06:00.930 --> 00:06:03.120 it? You have to maintain your own server, deal 00:06:03.120 --> 00:06:06.000 with security, manage downtime. Why would you 00:06:06.000 --> 00:06:08.420 choose that headache over the simple cloud version? 00:06:08.579 --> 00:06:11.250 Because without that self -hosted setup, You 00:06:11.250 --> 00:06:13.269 just can't run the video transformations you 00:06:13.269 --> 00:06:16.110 need. The trade -off is unavoidable. You accept 00:06:16.110 --> 00:06:18.889 the small headache of maintaining, say, a $10 00:06:18.889 --> 00:06:22.129 a month server to unlock the massive power of 00:06:22.129 --> 00:06:24.089 automated video assembly. So the architecture 00:06:24.089 --> 00:06:27.189 just demands that disk access. It demands it, 00:06:27.209 --> 00:06:29.910 and self -hosting is what provides it. The complexity 00:06:29.910 --> 00:06:32.490 is justified by the functionality you get. Right, 00:06:32.550 --> 00:06:35.069 the ability to process files locally. Exactly. 00:06:35.110 --> 00:06:37.189 So once you have that architecture in place, 00:06:37.269 --> 00:06:40.050 the next step is actually creating the presenter. 00:06:40.490 --> 00:06:43.769 the digital twin. And using a tool like OneVideo, 00:06:44.110 --> 00:06:46.209 it looks like you have two main options for making 00:06:46.209 --> 00:06:48.769 the avatar. That's right. Option one is you can 00:06:48.769 --> 00:06:50.689 just use your own image. That's great if you're 00:06:50.689 --> 00:06:52.949 trying to build an existing brand, but you have 00:06:52.949 --> 00:06:55.670 to follow the rules. A clear shot, good lighting, 00:06:55.790 --> 00:06:59.829 and ideally a 9 .16 aspect ratio. And option 00:06:59.829 --> 00:07:02.910 two. Option two is generating a totally synthetic 00:07:02.910 --> 00:07:05.689 avatar, like the guide's Emma example. Where 00:07:05.689 --> 00:07:07.970 you use really detailed prompt, right? Describing 00:07:07.970 --> 00:07:10.310 everything from hair color to the shirt, just 00:07:10.310 --> 00:07:12.509 to get a stable image that becomes the foundation 00:07:12.509 --> 00:07:14.730 of the avatar. And that initial generation is 00:07:14.730 --> 00:07:16.930 so cheap, it's about 50 cents. And once you have 00:07:16.930 --> 00:07:19.389 that perfect four second animated look, you can 00:07:19.389 --> 00:07:22.189 reuse it forever. But, you know, I'll admit I 00:07:22.189 --> 00:07:24.430 still wrestle with pump drift myself when I'm 00:07:24.430 --> 00:07:27.050 trying to optimize these synthetic images for 00:07:27.050 --> 00:07:30.610 consistency. Prompt version. Yeah, I mean. Even 00:07:30.610 --> 00:07:32.970 if you use the exact same input prompt, the AI 00:07:32.970 --> 00:07:35.430 might subtly change the lighting or a tiny facial 00:07:35.430 --> 00:07:37.850 expression, maybe the angle. And it can just 00:07:37.850 --> 00:07:40.589 slightly undermine the brand consistency that 00:07:40.589 --> 00:07:42.529 you need when you're doing high volume. It takes 00:07:42.529 --> 00:07:44.769 a lot of meticulous testing. That makes perfect 00:07:44.769 --> 00:07:46.970 sense. Consistency is everything at that scale. 00:07:47.050 --> 00:07:49.970 Right. So now, the voice. That's what really 00:07:49.970 --> 00:07:52.449 brings the avatar to life. And Eleven Lab seems 00:07:52.449 --> 00:07:55.720 to be the go -to tool for quality. Absolutely. 00:07:55.899 --> 00:07:58.160 You pick a voice based on your content style. 00:07:58.439 --> 00:08:00.660 You know, are you doing upbeat news? Maybe you 00:08:00.660 --> 00:08:03.319 pick Sally Ford. Is it serious financial commentary? 00:08:03.459 --> 00:08:05.779 Maybe Eve is a better fit. You just choose the 00:08:05.779 --> 00:08:08.740 voice ID, paste it into your Airtable, and that 00:08:08.740 --> 00:08:12.180 links that specific tone to your animated avatar 00:08:12.180 --> 00:08:15.060 clip. Okay, so strategically, forgetting the 00:08:15.060 --> 00:08:17.680 cost for a second, why is matching the voice 00:08:17.680 --> 00:08:20.860 tone so critical for an AI avatar? It's because 00:08:20.860 --> 00:08:23.180 the visual connection is already a little bit... 00:08:23.439 --> 00:08:26.139 artificial because of the AI generation. The 00:08:26.139 --> 00:08:28.480 voice becomes your primary way to establish authority 00:08:28.480 --> 00:08:31.779 or warmth or urgency. If the voice is flat, but 00:08:31.779 --> 00:08:34.240 the content is exciting, the viewer just drops 00:08:34.240 --> 00:08:36.100 off. It has to feel connected. But the voice's 00:08:36.100 --> 00:08:38.399 energy and tone have to perfectly align with 00:08:38.399 --> 00:08:40.879 the content style. Has to. So let's get into 00:08:40.879 --> 00:08:43.389 the step -by -step pipeline. The moment the engine 00:08:43.389 --> 00:08:46.029 starts, the system needs those viral content 00:08:46.029 --> 00:08:49.370 ideas, the fuel. It looks to X for high -performing 00:08:49.370 --> 00:08:51.710 recent videos. It's searching for things with 00:08:51.710 --> 00:08:54.450 100 ,000 views or more. This is just pure leverage. 00:08:54.769 --> 00:08:56.990 You are curating what's already proven to get 00:08:56.990 --> 00:09:00.389 engagement. And on top of that, when you feature 00:09:00.389 --> 00:09:02.570 a company's product, like in the guide's example 00:09:02.570 --> 00:09:04.809 of the Gemini 3 announcement, you're basically 00:09:04.809 --> 00:09:07.149 giving them free marketing. Which creates organic 00:09:07.149 --> 00:09:09.110 reach for you. Exactly. You're building your 00:09:09.110 --> 00:09:11.610 content on borrowed authority. Okay, so phase 00:09:11.610 --> 00:09:14.309 one starts when that viral link is found. It 00:09:14.309 --> 00:09:16.250 needs to scrape the info. This is where a tool 00:09:16.250 --> 00:09:19.590 like Appify comes in. Correct. Appify is essentially 00:09:19.590 --> 00:09:23.269 a smart digital agent. It scrapes specific data 00:09:23.269 --> 00:09:25.950 like video metadata and links from sites like 00:09:25.950 --> 00:09:28.649 X. And once it has that raw material, it feeds 00:09:28.649 --> 00:09:30.950 it into ChatGPT. Which is tasked with writing 00:09:30.950 --> 00:09:34.210 a short, high -impact script for short -form 00:09:34.210 --> 00:09:36.990 video. Yep. And that script immediately goes 00:09:36.990 --> 00:09:39.909 to the Eleven Labs engine in Phase 2, which creates 00:09:39.909 --> 00:09:41.909 that natural -sounding audio in the voice we 00:09:41.909 --> 00:09:44.289 already picked out. Then Phase 3 is the Avatar 00:09:44.289 --> 00:09:46.809 lip sync. The new audio gets injected into the 00:09:46.809 --> 00:09:49.029 video loop, and the lip sync model makes sure 00:09:49.029 --> 00:09:51.789 the mouth movements match the dialogue as believably 00:09:51.789 --> 00:09:54.070 as it can. Then you get to phase four, which 00:09:54.070 --> 00:09:56.570 is the assembly step. This is where the NE -10 00:09:56.570 --> 00:09:59.610 workflow stitches everything together. The talking 00:09:59.610 --> 00:10:02.629 avatar intro, the viral source clip, maybe some 00:10:02.629 --> 00:10:05.590 background music, the voiceover. It automatically 00:10:05.590 --> 00:10:08.070 resizes and crops everything for vertical video. 00:10:08.269 --> 00:10:11.620 And phase five is the polish. Whisper AI, the 00:10:11.620 --> 00:10:14.080 transcription tool, transcribes the audio and 00:10:14.080 --> 00:10:15.980 then it burns the captions right into the final 00:10:15.980 --> 00:10:19.039 video. So no manual editing for that crucial 00:10:19.039 --> 00:10:21.440 accessibility feature. This is where automation 00:10:21.440 --> 00:10:24.669 really shows its value. I thought the troubleshooting 00:10:24.669 --> 00:10:27.330 process they documented, the Emma test, was actually 00:10:27.330 --> 00:10:29.429 the best part. Such a great real -world insight. 00:10:29.710 --> 00:10:31.769 Oh, yeah. So the first test video they generated, 00:10:31.830 --> 00:10:33.669 the captions were just completely covering the 00:10:33.669 --> 00:10:35.789 avatar's face. Right. If you were doing that 00:10:35.789 --> 00:10:37.909 manually, that's five minutes of editing on every 00:10:37.909 --> 00:10:40.590 single video. But because it's a modular workflow, 00:10:40.750 --> 00:10:42.870 the fix was simple. They just adjusted the crop 00:10:42.870 --> 00:10:45.309 setting in NA10. I think they increased the top 00:10:45.309 --> 00:10:48.929 crop to 150 pixels. And that corrected the caption 00:10:48.929 --> 00:10:51.330 placement for every future video automatically. 00:10:51.370 --> 00:10:54.419 That is the true power of it. You fix the workflow 00:10:54.419 --> 00:10:57.000 one time and you've solved the problem for the 00:10:57.000 --> 00:11:00.320 next 150 videos without touching it again. The 00:11:00.320 --> 00:11:02.919 only human work is fixing that initial design 00:11:02.919 --> 00:11:05.659 flaw. So if the automation is that powerful, 00:11:05.840 --> 00:11:08.899 where is the single most critical moment? for 00:11:08.899 --> 00:11:10.740 making sure the quality is right. It's in the 00:11:10.740 --> 00:11:13.500 rigorous testing and precise adjustment of those 00:11:13.500 --> 00:11:16.279 workflow settings on the very first video. That 00:11:16.279 --> 00:11:19.059 prevents all the errors down the line. It's crucial 00:11:19.059 --> 00:11:21.000 to point out that this guide isn't really advising 00:11:21.000 --> 00:11:23.679 a total replacement of the human creator. This 00:11:23.679 --> 00:11:26.299 system is designed more as an augmentation tool, 00:11:26.460 --> 00:11:29.299 not a full substitute. Yeah, the strategic recommendation 00:11:29.299 --> 00:11:32.259 is a hybrid approach. Use the AI avatars for 00:11:32.259 --> 00:11:34.379 your high volume stuff for consistency, for daily 00:11:34.379 --> 00:11:37.399 news updates. That could easily be, say, 70 % 00:11:37.399 --> 00:11:39.639 of your output. But you still need to show up 00:11:39.639 --> 00:11:41.620 in person to build that deep connection, that 00:11:41.620 --> 00:11:43.740 trust, and to share emotional stories. That's 00:11:43.740 --> 00:11:46.759 the essential 30%. You're using AI for the tactical 00:11:46.759 --> 00:11:49.019 work, which frees you up for strategic work. 00:11:49.259 --> 00:11:52.379 And that human review layer is the absolute safeguard. 00:11:52.639 --> 00:11:56.019 The idea scraper feeds the ideas table every 00:11:56.019 --> 00:11:58.279 morning. The automation builds the videos. But 00:11:58.279 --> 00:12:00.779 you are still the final curator. Right. You watch 00:12:00.779 --> 00:12:03.320 the generated content. And only when you switch 00:12:03.320 --> 00:12:06.220 the status to schedule does the publishing automation. 00:12:07.120 --> 00:12:09.700 maybe using a tool like Blotato, take over. And 00:12:09.700 --> 00:12:12.080 that oversight is so important. It prevents brand 00:12:12.080 --> 00:12:15.840 damage from, you know, an inevitable AI mistake, 00:12:16.080 --> 00:12:18.720 a weird script, a bad caption, poor stitching. 00:12:18.940 --> 00:12:21.620 You get to maintain quality control even when 00:12:21.620 --> 00:12:23.820 you're working at scale. Okay, speaking of control, 00:12:23.899 --> 00:12:25.899 let me just push back a little here. This whole 00:12:25.899 --> 00:12:28.179 low -cost system relies on, what, five or six 00:12:28.179 --> 00:12:31.500 different third -party APIs, 11 labs, chat GPT, 00:12:31.580 --> 00:12:34.559 one video. Doesn't creating that many external 00:12:34.559 --> 00:12:36.519 dependencies make your business hugely vulnerable? 00:12:37.019 --> 00:12:39.019 I mean, what if one of them changes their pricing 00:12:39.019 --> 00:12:41.200 or just shuts down? It absolutely introduces 00:12:41.200 --> 00:12:44.139 vulnerability. But you accept that risk because 00:12:44.139 --> 00:12:46.279 the cost of trying to scale without these APIs 00:12:46.279 --> 00:12:48.820 is just exponentially higher. The strategy is 00:12:48.820 --> 00:12:51.480 to, one, choose vendors who are market leaders 00:12:51.480 --> 00:12:54.629 whose pricing is stable. And two, build your 00:12:54.629 --> 00:12:56.950 workflow so the components are swappable if you 00:12:56.950 --> 00:12:59.029 need to. You're really just betting that API 00:12:59.029 --> 00:13:01.009 access is only going to get cheaper and faster 00:13:01.009 --> 00:13:03.710 over time. That's a pretty powerful bet on technological 00:13:03.710 --> 00:13:06.509 progress. And looking ahead, I mean, the pace 00:13:06.509 --> 00:13:09.389 of change is just staggering. What does the roadmap 00:13:09.389 --> 00:13:12.580 tell us? Well, near term. So three to six months, 00:13:12.700 --> 00:13:14.740 we should expect really rapid quality upgrades. 00:13:14.860 --> 00:13:17.480 We're going to see near perfect lip sync, much 00:13:17.480 --> 00:13:20.080 more natural gestures and probably lower API 00:13:20.080 --> 00:13:23.039 costs as competition heats up. And medium term, 00:13:23.220 --> 00:13:25.740 six to 12 months out. Things get even crazier. 00:13:25.879 --> 00:13:27.580 We're looking at real -time video generation 00:13:27.580 --> 00:13:30.440 in seconds, not minutes, and truly interactive 00:13:30.440 --> 00:13:32.879 avatars that can actually reply to comments with 00:13:32.879 --> 00:13:35.399 newly generated video responses. Just think about 00:13:35.399 --> 00:13:37.740 the scale of knowledge transfer that unlocks. 00:13:37.919 --> 00:13:40.480 It completely changes how we interact with education, 00:13:40.779 --> 00:13:43.200 with customer service, everything online. Whoa, 00:13:43.379 --> 00:13:46.340 imagine scaling that to a billion queries, generating 00:13:46.340 --> 00:13:49.340 personalized, unique video responses instantly. 00:13:49.620 --> 00:13:52.799 That is a truly revolutionary shift. Yeah. So 00:13:52.799 --> 00:13:55.350 if we look at that future roadmap. What human 00:13:55.350 --> 00:13:57.450 function do you think will become completely 00:13:57.450 --> 00:14:00.210 obsolete for this kind of high volume content? 00:14:00.529 --> 00:14:02.710 The need for a person to maintain a physical 00:14:02.710 --> 00:14:05.590 on -camera presence for high volume educational 00:14:05.590 --> 00:14:08.289 content delivery. So what does all this really 00:14:08.289 --> 00:14:10.509 mean for an ambitious content creator today? 00:14:10.649 --> 00:14:13.330 This system is a fundamental shift in the economics 00:14:13.330 --> 00:14:15.909 of content. Your growth is no longer limited 00:14:15.909 --> 00:14:18.009 by the number of hours you can physically spend 00:14:18.009 --> 00:14:21.049 filming and scripting and editing. Exactly. The 00:14:21.049 --> 00:14:23.549 old model was all about expensive human hours. 00:14:23.820 --> 00:14:26.340 This new model is all about inexpensive, automated 00:14:26.340 --> 00:14:29.480 API calls. It allows a single creator to run 00:14:29.480 --> 00:14:31.840 a massive content studio, publishing multiple 00:14:31.840 --> 00:14:34.440 times a day across five or six platforms. And 00:14:34.440 --> 00:14:36.399 the cost transparency is what makes it so legitimate. 00:14:36.620 --> 00:14:39.100 We talked about maybe $11 a month for the creator 00:14:39.100 --> 00:14:42.480 plan at 11 labs, plus, what, $5 to $15 for self 00:14:42.480 --> 00:14:45.080 -hosting a scene? Yeah, and when you factor in 00:14:45.080 --> 00:14:48.019 all the necessary API costs for the scripts, 00:14:48.240 --> 00:14:50.480 the lip sync, the captions, the cloud hosting, 00:14:50.600 --> 00:14:53.320 the total monthly cost to run the whole... and 00:14:53.320 --> 00:14:57.840 generate 150 finished videos is about $145 a 00:14:57.840 --> 00:15:00.200 month. That cost is just incredibly compelling. 00:15:00.460 --> 00:15:02.279 You're getting results that would easily cost 00:15:02.279 --> 00:15:04.899 thousands of dollars if you outsourced it to 00:15:04.899 --> 00:15:07.580 human creators and editors. It's undeniable scale 00:15:07.580 --> 00:15:10.799 and efficiency. The revolution is here. And the 00:15:10.799 --> 00:15:12.879 creators who are already winning with this prove 00:15:12.879 --> 00:15:16.919 that value and utility, not biology, are what 00:15:16.919 --> 00:15:19.440 the audience really prioritizes. And that leads 00:15:19.440 --> 00:15:21.700 to a pretty profound question to leave you with. 00:15:22.090 --> 00:15:25.889 If audiences consistently prioritize pure informational 00:15:25.889 --> 00:15:29.149 value over the biological presence of a human 00:15:29.149 --> 00:15:31.950 presenter, how long until human -made content 00:15:31.950 --> 00:15:34.909 becomes the niche, the special artisanal choice, 00:15:35.070 --> 00:15:37.450 instead of the standard expectation? Something 00:15:37.450 --> 00:15:39.470 to ponder as you start sketching out the architecture 00:15:39.470 --> 00:15:42.009 for your own content machine. Thanks for joining 00:15:42.009 --> 00:15:43.429 us for this deep dive. We'll talk to you next 00:15:43.429 --> 00:15:43.649 time.